Continuation of #1309: Implementation of the resource manager server (issue #1294)

CHANGELOG.md

@@ -5,6 +5,28 @@ Versioned according to [Semantic Versioning](http://semver.org/).
 
 ## Unreleased
 
+Added:
+
+  * Resource Manager Server as `ocrd_network` analogon of `ocrd.cli.resmgr`, #1309
+    * `ocrd network resmgr-server` for triggering Resource Manager Server (RMS) in the background
+    * Processing Server also deploys RMS on each processing host
+    * ...
+
+Fixed:
+
+  * `ocrd.resource_manager`: ensure necessary + reduce unnecessary updates of user database
+  * `ocrd.resource_manager`: deduplicate entries (newer wins) before updating user database
+  * `ocrd resmgr download`: extract archives independent of whether they are URLs or local paths
+  * `ocrd resmgr download`: if `--overwrite`, ensure the old res gets removed
+  * `ocrd resmgr download`: default to `data` location instead of first in list of allowed
+  * `ocrd_utils.list_all_resources`: filter module non-resource files w/ more anti-patterns
+  * `ocrd_utils.list_all_resources`: no subpaths except for `cwd` location, OCR-D/spec#263, #1315
+  * `ocrd_utils.list_all_resources`: filter resources via media (MIME) type, if specified, #1315
+
+Removed:
+
+  * `ocrd resmgr download`: `*` (asterisk) as wildcard for all processors removed, processor must be provided explicitly, #1319
+
 ## [3.7.0] - 2025-11-02
 
 Changed:

requirements.txt

@@ -3,12 +3,13 @@ beanie~=1.7
 click >=7
 cryptography < 43.0.0
 Deprecated == 1.2.0
-docker
+docker>=7.1.0
 elementpath
 fastapi>=0.78.0
 filetype
 Flask
 frozendict>=2.4.0
+gitpython
 gdown
 httpx>=0.22.0
 importlib_metadata ; python_version < '3.8'

src/ocrd/cli/network.py

@@ -12,6 +12,7 @@
     client_cli,
     processing_server_cli,
     processing_worker_cli,
+    resource_manager_server_cli
 )
 
 
@@ -26,3 +27,4 @@ def network_cli():
 network_cli.add_command(client_cli)
 network_cli.add_command(processing_server_cli)
 network_cli.add_command(processing_worker_cli)
+network_cli.add_command(resource_manager_server_cli)

src/ocrd/cli/resmgr.py

@@ -20,6 +20,7 @@
     get_ocrd_tool_json,
     initLogging,
     RESOURCE_LOCATIONS,
+    RESOURCE_TYPES
 )
 from ocrd.constants import RESOURCE_USER_LIST_COMMENT
 
@@ -70,16 +71,16 @@
 @resmgr_cli.command('download')
 @click.option('-n', '--any-url', default='', help='URL of unregistered resource to download/copy from')
 @click.option('-D', '--no-dynamic', default=False, is_flag=True,
-              help="Whether to skip looking into each processor's --dump-{json,module-dir} for module-level resources")
-@click.option('-t', '--resource-type', type=click.Choice(['file', 'directory', 'archive']), default='file',
-              help='Type of resource',)
-@click.option('-P', '--path-in-archive', default='.', help='Path to extract in case of archive type')
+              help="Skip looking into each processor's --dump-{json,module-dir} module-registered resources")
+@click.option('-t', '--resource-type', type=click.Choice(RESOURCE_TYPES), default='file',
+              help='Type of resource (when unregistered or incomplete)',)
+@click.option('-P', '--path-in-archive', default='.', help='Path to extract in case of archive type (when unregistered or incomplete)')
 @click.option('-a', '--allow-uninstalled', is_flag=True,
-              help="Allow installing resources for uninstalled processors",)
+              help="Allow installing resources for not installed processors",)
 @click.option('-o', '--overwrite', help='Overwrite existing resources', is_flag=True)
-@click.option('-l', '--location', type=click.Choice(RESOURCE_LOCATIONS),
+@click.option('-l', '--location', type=click.Choice(RESOURCE_LOCATIONS), 
               help="Where to store resources - defaults to first location in processor's 'resource_locations' "
-                   "list or finally 'data'")
+                   "list, i.e. usually 'data'")
 @click.argument('executable', required=True)
 @click.argument('name', required=False)
 def download(any_url, no_dynamic, resource_type, path_in_archive, allow_uninstalled, overwrite, location, executable,
@@ -106,8 +107,6 @@
         executable = None
     if name == '*':
         name = None
-    is_url = (any_url.startswith('https://') or any_url.startswith('http://')) if any_url else False
-    is_filename = Path(any_url).exists() if any_url else False
     if executable and not which(executable):
         if not allow_uninstalled:
             log.error(f"Executable '{executable}' is not installed. "
@@ -126,65 +125,30 @@
                 'path_in_archive': path_in_archive}]
             )]
     for this_executable, this_reslist in reslist:
-        for resdict in this_reslist:
-            if 'size' in resdict:
-                registered = "registered"
-            else:
-                registered = "unregistered"
-            if any_url:
-                resdict['url'] = any_url
-            if resdict['url'] == '???':
-                log.warning(f"Cannot download user resource {resdict['name']}")
-                continue
-            if resdict['url'].startswith('https://') or resdict['url'].startswith('http://'):
-                log.info(f"Downloading {registered} resource '{resdict['name']}' ({resdict['url']})")
-                if 'size' not in resdict:
-                    with requests.head(resdict['url']) as r:
-                        resdict['size'] = int(r.headers.get('content-length', 0))
-            else:
-                log.info(f"Copying {registered} resource '{resdict['name']}' ({resdict['url']})")
-                urlpath = Path(resdict['url'])
-                resdict['url'] = str(urlpath.resolve())
-                if Path(urlpath).is_dir():
-                    resdict['size'] = directory_size(urlpath)
-                else:
-                    resdict['size'] = urlpath.stat().st_size
-            if not location:
-                location = get_ocrd_tool_json(this_executable)['resource_locations'][0]
-            elif location not in get_ocrd_tool_json(this_executable)['resource_locations']:
-                log.error(f"The selected --location {location} is not in the {this_executable}'s resource search path, "
-                          f"refusing to install to invalid location")
-                sys.exit(1)
-            if location != 'module':
-                basedir = resmgr.location_to_resource_dir(location)
-            else:
-                basedir = get_moduledir(this_executable)
-                if not basedir:
-                    basedir = resmgr.location_to_resource_dir('data')
-
+        resource_locations = get_ocrd_tool_json(this_executable)['resource_locations']
+        if not location:
+            location = resource_locations[0]
+        elif location not in resource_locations:
+            log.warning(f"The selected --location {location} is not in the {this_executable}'s resource search path, "
+                        f"refusing to install to invalid location. Instead installing to: {resource_locations[0]}")
+        res_dest_dir = resmgr.build_resource_dest_dir(location=location, executable=this_executable)
+        for res_dict in this_reslist:
             try:
-                with click.progressbar(length=resdict['size']) as bar:
-                    fpath = resmgr.download(
-                        this_executable,
-                        resdict['url'],
-                        basedir,
-                        name=resdict['name'],
-                        resource_type=resdict.get('type', resource_type),
-                        path_in_archive=resdict.get('path_in_archive', path_in_archive),
-                        overwrite=overwrite,
-                        no_subdir=location in ['cwd', 'module'],
-                        progress_cb=lambda delta: bar.update(delta)
-                    )
-                if registered == 'unregistered':
-                    log.info(f"{this_executable} resource '{name}' ({any_url}) not a known resource, creating stub "
-                             f"in {resmgr.user_list}'")
-                    resmgr.add_to_user_database(this_executable, fpath, url=any_url)
-                resmgr.save_user_list()
-                log.info(f"Installed resource {resdict['url']} under {fpath}")
+                fpath = resmgr.handle_resource(
+                    res_dict=res_dict,
+                    executable=this_executable,
+                    dest_dir=res_dest_dir,
+                    any_url=any_url,
+                    overwrite=overwrite,
+                    resource_type=resource_type,
+                    path_in_archive=path_in_archive
+                )
+                if not fpath:
+                    continue
             except FileExistsError as exc:
                 log.info(str(exc))
-            log.info(f"Use in parameters as "
-                     f"'{resmgr.parameter_usage(resdict['name'], usage=resdict.get('parameter_usage', 'as-is'))}'")
+            usage = res_dict.get('parameter_usage', 'as-is')
+            log.info(f"Use in parameters as '{resmgr.parameter_usage(res_dict['name'], usage)}'")
 
 
 @resmgr_cli.command('migrate')

src/ocrd/constants.py

@@ -9,7 +9,6 @@
     'DOWNLOAD_DIR',
     'DEFAULT_REPOSITORY_URL',
     'BASHLIB_FILENAME',
-    'RESOURCE_LIST_FILENAME',
     'BACKUP_DIR',
     'RESOURCE_USER_LIST_COMMENT',
 ]
@@ -19,6 +18,5 @@
 DOWNLOAD_DIR = '/tmp/ocrd-core-downloads'
 DEFAULT_REPOSITORY_URL = 'http://localhost:5000/'
 BASHLIB_FILENAME = resource_filename(__package__, 'lib.bash')
-RESOURCE_LIST_FILENAME = resource_filename(__package__, 'resource_list.yml')
 RESOURCE_USER_LIST_COMMENT = "# OCR-D private resource list (consider sending a PR with your own resources to OCR-D/core)"
 BACKUP_DIR = '.backup'

src/ocrd/processor/base.py

@@ -42,15 +42,14 @@
 from ocrd_utils import (
     VERSION as OCRD_VERSION,
     MIMETYPE_PAGE,
-    MIME_TO_EXT,
     config,
     getLogger,
     list_resource_candidates,
-    pushd_popd,
     list_all_resources,
     get_processor_resource_types,
     resource_filename,
     parse_json_file_with_comments,
+    pushd_popd,
     make_file_id,
     deprecation_warning
 )
@@ -555,9 +554,9 @@
                     self._base_logger.debug("stopped executor %s after %d tasks", str(executor), len(tasks) if tasks else -1)
                     if max_workers > 1:
                         # can cause deadlock:
                         #log_listener.stop()
                         # not much better:
                         #log_listener.enqueue_sentinel()
                         pass
 
             except NotImplementedError:
@@ -640,7 +639,7 @@
                 self._base_logger.error(repr(e))
                 self._base_logger.warning(f"failed downloading file {input_file} for page {page_id}")
         # process page
         #executor.submit(self.process_page_file, *input_files)
         return executor.submit(_page_worker, max_seconds, *input_files), page_id, input_files
 
     def process_workspace_handle_tasks(self, tasks: Dict[TFuture, Tuple[str, List[Optional[OcrdFileType]]]]) -> Tuple[
@@ -915,7 +914,7 @@
                                          value=self.version),
                                LabelType(type_='ocrd/core',
                                          value=OCRD_VERSION)])
             ])
         metadata_obj.add_MetadataItem(metadata_item)
 
     def resolve_resource(self, val):
@@ -935,9 +934,8 @@
             cwd = self.old_pwd
         else:
             cwd = getcwd()
-        ret = [cand for cand in list_resource_candidates(executable, val,
-                                                         cwd=cwd, moduled=self.moduledir)
-               if exists(cand)]
+        ret = list(filter(exists, list_resource_candidates(executable, val,
+                                                           cwd=cwd, moduled=self.moduledir)))
         if ret:
             self._base_logger.debug("Resolved %s to absolute path %s" % (val, ret[0]))
             return ret[0]
@@ -968,17 +966,9 @@
         """
         List all resources found in the filesystem and matching content-type by filename suffix
         """
-        mimetypes = get_processor_resource_types(None, self.ocrd_tool)
-        for res in list_all_resources(self.ocrd_tool['executable'], moduled=self.moduledir):
+        for res in list_all_resources(self.executable, ocrd_tool=self.ocrd_tool, moduled=self.moduledir):
             res = Path(res)
-            if '*/*' not in mimetypes:
-                if res.is_dir() and 'text/directory' not in mimetypes:
-                    continue
-                # if we do not know all MIME types, then keep the file, otherwise require suffix match
-                if res.is_file() and not any(res.suffix == MIME_TO_EXT.get(mime, res.suffix)
-                                             for mime in mimetypes):
-                    continue
-            yield res
+            yield res.name
 
     @property
     def module(self):
@@ -1193,7 +1183,7 @@
                     if hasattr(file, 'pageId')), "")
     pool = ThreadPool(processes=1)
     try:
         #_page_worker_processor.process_page_file(*input_files)
         async_result = pool.apply_async(_page_worker_processor.process_page_file, input_files)
         async_result.get(timeout or None)
         _page_worker_processor.logger.debug("page worker completed for page %s", page_id)